OMRON SINIC X to Present Four Research Papers at a Top-Tier Conference on Robotics

• May 29, 2023

OMRON SINIC X Corporation (HQ: Bunkyo-ku, Tokyo; President and CEO: Masaki Suwa; hereinafter “OSX”) is pleased to announce that four of its research papers have been accepted for publication at the 2023 IEEE International Conference on Robots and Automation (hereinafter “ICRA 2023”), a leading international event in the field. The papers will be presented at ICRA 2023, which is held in London, the United Kingdom, from May 29, 2023.

ICRA is one of the largest and most influential international conferences on robotics and automation. For its 2023 event, 43.04% of the 3,125 research papers submitted have been accepted.
OSX’s research goal is to create a “future where humans and machines collaborate in harmony,” with machines ably assisting humans with a good understanding of the intent of work to enhance human creativity. Accordingly, in addition to internship cooperation, OSX proactively conducts joint research not only with the OMRON Group but also with external partners for research robots that move flexibly like humans or autonomously cooperate with other robots. For ICRA 2023, four research papers have been accepted in recognition of its research findings in the field of robotics technology.
These research findings are released as open-source code in articles with a simple explanation about technology and on GitHub so that it can be actively used toward social implementation. For details, please confirm the link of each paper.

OSX continue to develop value creation via technological innovation through collaboration with universities and external research institutions.

<Publication> *All presentation times are in local time.
Learning Food Picking without Food: Fracture Anticipation by Breaking Reusable Fragile Objects

Authors	Rinto Yagawa (Keio University), Reina Ishikawa (Keio University), Masashi Hamaya (OSX), Kazutoshi Tanaka (OSX), Atsushi Hashimoto (OSX), Hideo Saito (Keio University)
Time of presentation	08:30–10:10, May 30, 2023
Overview	Food picking is trivial for humans but not for robots, as foods are fragile. Presetting foods' physical properties does not help robots much due to the objects' inter- and intra- category diversity. In a previous study, we demonstrated that learning food properties is essential for fracture-free picking. We leveraged the breaking experiences of food objects to obtain food properties. During the training, a robot equipped with a universal two-finger gripper and tactile sensors broke the objects and measured the tactile signals, which describe when the fractures occur. We then trained a fracture anticipation network, which anticipates the timing of the object's future fracture. Fracture anticipation enables fracture-free picking by stopping the gripper motion immediately before the object breaks. The previous method clarified that we could achieve food picking through object-breaking experiences; however, the method assumes a category-specific model and requires breaking many food objects in the category to train the model. For practical use, the network should be able to anticipate the fracture of unseen food objects. Furthermore, the amount of food consumption required for training is economically and environmentally undesirable. In this paper, we propose a method for making a robot grasp multiple unseen food categories without using a category-specific model trained by consuming a considerable amount of food in that category. The key idea is to leverage the object-breaking experiences of several reusable fragile objects instead of consuming real foods while making the picking ability object-invariant with domain generalization (DG). In real-robot experiments, we trained a model with reusable objects (toy blocks, ping-pong balls, and jellies), selected based on the three common fracture types (crack, rupture, and crush). We then tested the model with four real food objects (tofu, bananas, potato chips, and tomatoes). The results showed that the proposed combination of reusable objects' breaking experiences and DG is effective for the food-picking task.
Details	https://omron-sinicx.github.io/mgrasp/ https://youtu.be/9cW1KVdGHmY

ViewBirdiformer: Learning to recover ground-plane crowd trajectories and ego-motion from a single ego-centric view

Authors	Mai Nishimura (OSX), Shohei Nobuhara (Kyoto University), Ko Nishino (Kyoto University)
Time of presentation	15:00–16:40, May 30, 2023
Overview	Self-localization in densely crowded environments has been a long-standing challenge in the fields of computer vision and robotics. Conventional localization methods including SLAM (Simultaneous Localization and Mapping) rely on the static world assumption, where an observer can robustly detect and track static keypoints across the frames. However, in dynamic environments, these approaches often fail due to frequent occlusions caused by pedestrians. In this study, we propose ViewBirdiformer, a novel Transformer-based network that learns to recover the on-ground trajectories of the observation camera and its surrounding pedestrians solely from perceived movements in an ego-centric view. ViewBirdiformer achieves real-time localization in densely crowded environments with a single inference pass, which opens a new avenue of real-time applications.
Details	https://www.youtube.com/watch?v=PmQV8-Iz9Qg https://arxiv.org/abs/2210.06332

Twist Snake: Plastic table-top cable-driven robotic arm with all motors located at the base link

Authors	Kazutoshi Tanaka (OSX), Masashi Hamaya (OSX)
Time of presentation	15:00–16:40, May 31, 2023
Overview	Table-top robotic arms have been developed in research and are commercially available. These arms must be lightweight for safety. Therefore, this study aims to design a lightweight table-top robotic arm. We adopt a cable-driven mechanism used for lightweight robots. Particularly, we focus on locating all motors at the base link and using plastic structural parts to decrease the weight of a cable-driven robotic arm more. However, locating all motors at the base link results in a significant distance between a driving motor and driven joint, increases the number of parts for the force transmission. Using plastic structural parts increases the risk of a cable loosening and coming off of a pulley. To overcome these issues, this study proposed a novel cable-driven robotic arm named Twist Snake. We designed a joint composition of Twist Snake to minimize the number of parts for the force transmission. In addition, it has a compact cable-pretension/termination-mechanism and covering parts to prevent the cable from loosening and coming off of the pulley. The arm comprised 475 mm long moving links with an 802 g. The feasibility of the arm was experimentally demonstrated by contact-rich tasks, the insertion of a toy peg into a hole, and swiping a whiteboard with a cleaner.
Details	https://omron-sinicx.github.io/twistsnake/ https://youtu.be/sjfoP8uSgYY

Risk-aware Path Planning via Probabilistic Fusion of Traversability Prediction for Planetary Rovers on Heterogeneous Terrains

Authors	Masafumi Endo (Keio University; OSX intern from June 2022 to March 2023), Tatsunori Taniai (OSX), Ryo Yonetani (CyberAgent, Inc., [OSX at the time of writing), Genya Ishigami (Keio University)
Time of presentation	15:00–16:40, June 1, 2023
Overview	The autonomous navigation of rovers in planetary exploration poses challenges in accurately predicting traversability, particularly in heterogeneous deformable terrains with varying geological features. Machine learning (ML) techniques inevitably suffer from the risk of prediction errors, which can result in wheel slip and jeopardize mission success. To address this issue, this research introduces a novel path planning algorithm that explicitly considers erroneous prediction. The algorithm leverages the probabilistic fusion of distinctive ML models for terrain type classification and slip prediction, creating a unified distribution that captures multiple slip modes. This enables a more precise assessment of traversability on heterogeneous terrains. Further, this approach facilitates the application of statistical risk assessment techniques to derive risk-aware traversing costs for path planning. Comparative evaluations against existing methods demonstrate the algorithm's efficacy in reducing the risk of rover immobilization.
Details	https://omron-sinicx.github.io/safe-rover-navi/ https://www.youtube.com/watch?v=ax2SNa8vJ0k Risk-aware path planning method for mobile robots in planetary environments (ICRA 2023)

About OMRON SINIC X Corporation
OMRON SINIC X Corporation is a strategic subsidiary seeking to realize the “near future designs” that OMRON forecasts. Researchers with cutting-edge knowledge and experience across many technological domains, including AI, Robotics, IoT, and sensing, are affiliated with OSX, and with the aim of solving social issues, they are working to create near future designs by integrating innovative technologies with business models and strategies in technology and IP. The company will also accelerate the creation of near future designs through joint research with universities and external research institutions. For more details, please refer to https://www.omron.com/sinicx/en/